Bridging Specified States With Stochastic Behavioral-Consistent Vehicle Trajectories for Enhanced Digital Twin Simulation Realism

Hongsheng Qi, Chenxi Chen, Xianbiao Hu

Research output: Contribution to journalArticlepeer-review

Abstract

Digital twin (DT) technology integrates the physical world with its digitalized counterpart and suggests significant potential for intelligent transportation system development, such as CAV test and development. In the foreseeable near future, human-driven vehicles (HDVs) will continue to predominate, and a digital replica of the transportation system should reflect their behavioral patterns for enhanced simulation realism purposes. As such, stochastic driver behavior and vehicle dynamics should be respected. The observations serving as DT input, often captured at discrete moments (e.g., the roadside units and cameras are only installed at certain locations), result in discontinuously captured vehicle trajectories. The stochastic generation of behaviorally consistent vehicle trajectories conditional on such incomplete information becomes important. Current conditional approaches include modified Brownian bridge (MBB) and guided proposal bridge (GPB) may not be able to output realistic results. To fill this gap, we propose conditional generation methods of behaviorally consistent trajectories, employing the stochastic bridge approach for the first time. First, a vehicular dynamics model that encapsulates the stochasticity of the human-vehicle system is employed, and then we prove that MBB and GPB fail to generate satisfactory results. Then, a forward-backward method is proposed based on the backward Markov process, which takes the vehicular dynamics model as behavioral input. The proposed method is validated against real-world data and mainstream simulation platforms, showing that the forward-backward generation method provides consistent and realistic results. Its time consumption has also been proven to be promising for real-time DT applications.

Original languageEnglish (US)
Pages (from-to)18635-18646
Number of pages12
JournalIEEE Internet of Things Journal
Volume11
Issue number10
DOIs
StatePublished - May 15 2024

All Science Journal Classification (ASJC) codes

  • Signal Processing
  • Information Systems
  • Hardware and Architecture
  • Computer Science Applications
  • Computer Networks and Communications

Cite this